PositiveControl

Rails Rocket Sheep — AI skill for Claude Code

AI community

Rails 8 application template for building with AI coding agents (Claude Code, Cursor, Codex): generated CLAUDE.md, one-rule-per-file conventions, gates in CI.

How to install Rails Rocket Sheep

This entry records only its repository, not the path inside it, so there is no exact command to give. Open PositiveControl/rails-rocket-sheep and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Rails Rocket Sheep does

Rails 8 application template for building with AI coding agents (Claude Code, Cursor, Codex): generated CLAUDE.md, one-rule-per-file conventions, gates in CI.

Alternatives in AI

Vibe Coding Toolkit — A curated, battle-tested AI-coding toolkit: Claude Code plugins, subagent orchestration, quality gates, and re 297 ★
Zero Limit — ZeroLimit is a cross-platform desktop application for monitoring AI coding assistant quotas using CLIProxyAPI 234 ★
Rails AI Context — 45 MCP tools that give AI coding agents ground truth about your Rails app: schema, models, routes, controllers 155 ★

README

Rails Rocket Sheep 🚀🐑

**A Rails 8 application template for building with AI coding agents: Claude Code, Cursor, Codex, or any agent that reads a `CLAUDE.md` or `AGENTS.md`.**

`rails new` gives you a framework. Rocket Sheep gives you a codebase an agent can be productive in on the first prompt: the patterns are already there, the conventions are written down one rule per file, and a generated `CLAUDE.md` routes the agent to the right rule before it touches anything, and the rules a machine can check are checked at push time and in CI. The result is that the tenth feature looks like the first one.

It is applied once and leaves behind ordinary Rails files you own. There is no gem, no runtime dependency, and nothing to upgrade. MIT licensed.

rails new myapp --database=postgresql \
  --template=https://raw.githubusercontent.com/PositiveControl/rails-rocket-sheep/main/template.rb

What it adds to `rails new`

Conventions an agent reads. CLAUDE.md, AGENTS.md, and docs/rules/ in the generated app, one convention per file with the right and wrong version side by side, and a two-tier index that routes a file path or a symptom to the rule that governs it.
Patterns before the agent arrives. ApplicationService with a Result type, ApplicationForm, ApplicationComponent, Data-based registries, and a fixed budget of directories under app/ so the catalogue names its own limit.
A workflow, not just advice. 24 slash commands from "what should I do next" to a merged PR, with review gates, mirrored to Claude Code and Cursor.
Gates that do not depend on the agent cooperating. An agent will eventually miss a rule, so the rules a machine can check are checked. bin/gates runs from a git pre-push hook and again in CI, for every editor and agent, and refuses rejected patterns, a directory outside the app/ budget, a drifted command mirror, and a foreign key with no index. A query ledger fails CI on any